Surveys

           (Surveys and Sampling) Surveys

A survey is a series of self-report measures administered either through an interview or a written questionnaire. Surveys are the most widely used method of collecting descriptive information about a group of people. You may have received a phone call (it usually arrives in the middle of the dinner hour when most people are home) from a survey research group asking you about your taste in music, your shopping habits, or your political preferences. 

The goal of a survey, as with all descriptive research, is to produce a “snapshot” of the opinions, attitudes, or behaviors of a group of people at a given time. Because surveys can be used to gather information about a wide variety of information in a relatively short time, they are used extensively by businesspeople, advertisers, and politicians to help them learn what people think, feel, or do. 

Interviews

       Surveys are usually administered in the form of an interview, in which questions are read to the respondent in person or over the telephone. One advantage of in-person interviews is that they may allow the researcher to develop a close rapport and sense of trust with the respondent. This may motivate the respondent to continue with the interview and may lead to more honest and open responding. However, face-to-face interviews are extremely expensive to conduct, and consequently telephone surveys are now more common. In a telephone interview all of the interviewers are located in one place, the telephone numbers are generated automatically, and the questions are read from computer terminals in front of the researchers. This procedure provides such effi ciency and coordination among the interviewers that many surveys can be conducted in one day. 

Unstructured Interviews. Interviews may use either free-format or fixedformat self-report measures. In an unstructured interview the interviewer talks freely with the person being interviewed about many topics. Although a general list of the topics of interest is prepared beforehand, the actual interview focuses in on those topics that the respondent is most interested in or most knowledgeable about. Because the questions asked in an unstructured interview differ from respondent to respondent, the interviewer must be trained to ask questions in a way that gets the most information from the respondent and allows the respondent to express his or her true feelings. One type of a face-to-face unstructured interview in which a number of people are interviewed at the same time and share ideas both with the interviewer and with each other is called a focus group.

        Unstructured interviews may provide in-depth information about the particular concerns of an individual or a group of people, and thus, may produce ideas for future research projects or for policy decisions. It is, however, very diffi cult to adequately train interviewers to ask questions in an unbiased manner and to be sure that they have actually done so. And, as we have seen in Chapter 4, because the topics of conversation and the types of answers given in free-response formats vary across participants, the data are diffi cult to objectively quantify and analyze, and are therefore frequently treated qualitatively. 

Structured Interviews. Because researchers usually want more objective data, the structured interview, which uses quantitative fi xed-format items, is most common. The questions are prepared ahead of time, and the interviewer reads the questions to the respondent. The structured interview has the advantage over an unstructured interview of allowing better comparisons of the responses across different individuals because the questions, time frame, and response format are controlled to be the same for each respondent. 

Questionnaires

A questionnaire is a set of fi xed-format, self-report items that is completed by respondents at their own pace, often without supervision. Questionnaires are generally cheaper than interviews because a researcher can mail the questionnaires to many people or have them complete the questionnaires in large groups. Questionnaires may also produce more honest responses than interviews, particularly when the questions involve sensitive issues such as sexual activity or annual income, because respondents are more likely to perceive their responses as being anonymous than they are in interviews. In comparison to interviews, questionnaires are also likely to be less infl uenced by the characteristics of the experimenter. For instance, if the topic concerns race-related attitudes, how the respondent answers might depend on the race of the interviewer and how the respondent thinks the interviewer wants him or her to respond. Because the experimenter is not present when a questionnaire is completed, or at least is not directly asking the questions, such problems are less likely.

The Response Rate. Questionnaires are free of some problems that may occur in interviews, but they do have their own set of diffi culties. Although people may be likely to return surveys that have direct relevance to them (for instance, a survey of college students conducted by their own university), when mailings are sent to the general population, the response rate (that is, the percentage of people who actually complete the questionnaire and return it to the investigator) may not be very high. This may lead to incorrect conclusions because the people who return the questionnaire may respond differently than those who don’t return it would have. Investigators can sometimes increase response rates by providing gifts or monetary payments for completing the survey, by making the questionnaire appear brief and interesting, by ensuring the confi dentiality of all of the data, and by emphasizing the importance of the individual in the research (Dillman, 1978). Follow-up mailings can also be used to remind people that they have not completed the questionnaire, with the hope that they will then do so. 

Question Order. Another potential problem with questionnaires that does not occur with interviews is that people may not answer the questions in the order they are written, and the researcher does not know whether or not they have. To take one example, consider these two questions:

1. “How satisfied are you with your relationships with your family?”

2. “How satisfied are you with your relationship with your spouse?”

If the questions are answered in the order that they are presented here, then most respondents interpret the word family in question 1 to include their spouse. If question 2 is answered before question 1, however, the term family in question 1 is interpreted to mean the rest of the family except the spouse. Such variability can create measurement error (Schuman & Presser, 1981; Schwarz & Strack, 1991). 

Use of Existing Survey Data

 Because it is very expensive to conduct surveys, scientists often work together on them. For instance, a researcher may have a small number of questions relevant to his or her research included within a larger survey. Or researchers can access public-domain data sets that contain data from previous surveys. The U.S. Census is probably the largest such data set, containing information on family size, fertility, occupation, and income for the entire U.S. population, as well as a more extensive interview data set of a smaller group of citizens. The General Social Survey is a collection of over 1,000 items given to a sample of U.S. citizens (Davis, Smith, & Marsden, 2000). Because the same questions are asked each year the survey is given, comparisons can be made over time. Sometimes these data sets are given in comparable forms to citizens of different countries, allowing cross-cultural comparisons. One such data set is the Human Area Relations Files. Indexes of some of the most important social science databases can be found in Clubb, Austin, Geda, and Traugott (1985).